Chord: Open-Source Prototype for PBR Material Estimation Debuting at Siggraph (www.ubisoft.com)

Ubisoft La Forge has unveiled an open-source prototype for a new material estimation model called CHORD, recently accepted at SIGGRAPH Asia 2025. The model addresses the complex and time-consuming process of creating Physically Based Rendering (PBR) materials for 3D objects in video games, an area that traditionally requires extensive expertise in 3D software. By leveraging advances in AI, CHORD aims to streamline workflows for texture artists by generating high-quality base materials from text or image inputs through a multi-stage AI pipeline. This groundbreaking approach combines texture synthesis, material estimation, and upscaling, ultimately aiming to make the material creation process more accessible and efficient. CHORD employs innovative techniques like chained decomposition and LEGO-conditioning to enhance material map prediction from texture images. With its effective single-step training approach and usage of a large open-source dataset, the model has shown remarkable performance in generating multi-modal PBR material maps. While the results are promising, there are still quality improvements to be made, particularly for high-specular materials like metals. As La Forge embraces open-source contributions, CHORD represents a significant step towards empowering creators by integrating AI-assisted tools into existing workflows, potentially reshaping the landscape of game material production.